Process your first document
Start with one paper whose PDF is already available locally in Zotero. Processing uploads that PDF to the parser you configure; check your provider's terms before using confidential material.
1. Configure a parser
Open Zotero Settings → Knowledge MCP → Document parsing. Choose PaddleOCR (PaddleOCR VL-1.6) or MinerU (VLM), enter the matching service credentials, enable parsing and save. See service configuration for endpoint details.
2. Add the paper
In a native Zotero literature view, right-click the paper and use Add to Knowledge Base, selecting the desired knowledge library or folder. The plugin admits its metadata and queues eligible PDF attachments. Adding the same paper to the same location again does not create a duplicate resource.
An item without a PDF can still have metadata in the knowledge base, but it cannot acquire extracted text from that metadata alone.
3. Follow processing
Open Import Tasks. The workflow groups three stages:
- Add to knowledge base.
- Extract document content.
- Prepare semantic search, when enabled.
Extraction includes preparation, upload, provider waiting, extraction, download and local saving. A provider page percentage describes extraction progress, not the percentage of the complete workflow. You can inspect failures and use the available cancel/retry controls. Retrying the vector stage does not upload the PDF again.
4. Read the result
Open the PDF in Zotero's native reader and click M to open its Markdown companion. If the reader was open before processing completed, close and reopen it. Use the outline, document search and citation previews to navigate. See reading and editing.
5. Search or connect a client
Text search works with published text without an embedding service. For semantic search, configure and enable model services, then allow indexing to finish. Use search or connect a local AI client.